Modeling Of Multi-Dimensional Wind Farms Output Correlation Based On Mixture Vine Copula Structures And Its Application In Reactive Power Optimization

With the aggravating energy crisis and deteriorating environmental pollution problem, exploitation and utilization of clean energy to alleviate the awkward situation of energy and environment is becoming the focus of the public。In recent years, new energy generation has developed vigorously in our country, among which hydroelectric and wind power generation enjoy dominant positions. Hydroelectric is of good controllability for its general generation form of building dams, hence it could be used for peak frequency regulation besides mitigation of energy and environmental problems. But for wind power generation, its output power is largely dependent on wind speed, which holds strong random and intermittent features, brought in significant influences on power flow and system voltage when connected to power grid. Simultaneously, for adjacent relationship in position among different wind farms, output power of wind farms always show strong relevance. Aimed at the randomness and correlation of wind power output, a mixed vine copula model was proposed to build the correlation model of multi-dimension speed, and furtherly adopted to analysis its influences to power system operation, based on which a reactive power optimization model was build and reactive optimization was operated on power system with multiple wind farms integrated.A mixture model was proposed combining clustering method and vine Copula structure to describe the relevance among outputs of multiple wind farms. Clustering methods could classify data into groups according to similarities among different individuals, so as to analysis different characteristics in different groups. The function of vine copula structure is to turn the multi-dimension correlation problem into multiple problems of two-dimensional relevance, considering asymmetric correlation among random variables while reducing complexity of analyzing multi-dimension The combination of two methods offers a thorough way to consider correlation among output of wind farms in different scenarios, ensures a more close description of the real correlation between output of wind farms. To validate the rationality and effectiveness of the proposed strategy, three wind farms are integrated into IEEE 30 power system, based on the recorded data of wind farm output in someplace. The proposed strategy was adopted to build the correlation model and then samples was obtained according to ratios of numbers of individuals of each scenario and converged into one sample, which are later introduced in the aforementioned system to calculate probabilistic power flow. The comparison of probabilistic power flow results of sample data and record data proved that the proposed strategy could describe the correlation among wind farms output well.Later, the proposed strategy was combined with traditional static reactive power optimization model to build a probabilistic optimization model that consider the randomness of wind generation and correlation among outputs of multiple wind farms and modified a newly proposed intelligent algorithm called Backtracking Search Algorithm (BSA) to find the optimal solution of the probabilistic optimization model that aimed at improving long term economy and safety. At last, the modified intelligent algorithm was adopted in the aforementioned IEEE 30 system with three wind farms integrated to verify the feasibility of the modified algorithm for reactive power optimization problems.